How Baidu's business is powered by Deep Learning & AI

Deep learning is a technique to make computers learn using algorithms resulting in an artificial neural network. This requires considerable computing power and large data volumes to help algorithms accurately mine deep underlying relationships.

Baidu is the search leader in a country with a base of 700 million internet users. This provides it with the requisite data, and a clear advantage. Here are three ways Baidu is harnessing deep learning to power its business:

Deep Learning and AI for improving user engagement
  • Voice Recognition: Baidu is using AI to deliver a better user experience and improve customer satisfaction by making its apps more useful and efficient, beyond optimizing its flagship search. Chinese writing uses thousands of characters, which makes internet searches on a smartphone extremely impractical. Most consumers in China prefer the ease of voice recognition for the task, and Baidu claims a 97% success rate in the field. Its Deep Speech 2 algorithm can, in some cases, recognize English and Mandarin better than people. 
  • Translations: Baidu Translate, based on deep learning, can translate 28 languages and processes, on average, over 100 million requests a day. 
  • Customized suggestions / recommendations: Baidu is also using its AI-driven data to suggest customized restaurant and retail suggestions in its online-to-offline platform Nuomi. The data more accurately predicts traffic conditions for users of Baidu Maps, and provides more relevant content for the company's recently launched news feed.
Deep Learning and AI for improving business efficiency
  • Voice recognition based CRM: Baidu uses a voice-recognition-based CRM (customer relationship management) system to field customer questions, and provides sales staff with real-time answers and data. 
  • Facial recognition based entry-control: It is also testing a facial-recognition-based entry-control system at Baidu Campus in Beijing. 
  • Delivery Estimates: As Baidu Delivery service (formerly Baidu Takeout Delivery) expands from restaurants to other areas such as fresh-food delivery and on-demand logistics, AI provides better estimates for delivery times. 
Deep Learning and AI to pursue new business lines
  • AI-based financial risk control services : Baidu recently entered the financial-services market. Here, it uses AI-based risk-control techniques including facial and fingerprint recognition, liveliness detection, and optical character recognition of identification documents.
  • Autonomous driving: Baidu sees autonomous driving as an important area for future growth. It began testing driverless cars both in China and in California earlier this year. These cars use image recognition and deep-learning-based processing to navigate. Baidu has plans for small-scale mass production of autonomous vehicles by 2018.